SPACER: server for predicting allosteric communication and effects of regulation

Alexander Goncearenco 3 Simon Mitternacht 2 Taipang Yong 0 Birgit Eisenhaber 0 Frank Eisenhaber 0 5 6 7 Igor N. Berezovsky 4 7 0 Bioinformatics Institute (BII), Agency for Science, Technology and Research (A 1 07-01, Matrix, 138671, Singapore 2 University Library, University of Bergen , Bergen 5020, Norway 3 Computational Biology Unit and Department of Informatics, University of Bergen , Bergen 5020, Norway 4 Department of Biological Chemistry, Weizmann Institute of Science , Rehovot 76100, Israel 5 School of Computer Engineering (SCE), Nanyang Technological University (NTU) , 50 Nanyang Drive, 637553, Singapore 6 Department of Biological Sciences (DBS), National University of Singapore (NUS), 8 Medical Drive , 117597, Singapore 7 STAR), 30 Biopolis Street The SPACER server provides an interactive framework for exploring allosteric communication in proteins with different sizes, degrees of oligomerization and function. SPACER uses recently developed theoretical concepts based on the thermodynamic view of allostery. It proposes easily tractable and meaningful measures that allow users to analyze the effect of ligand binding on the intrinsic protein dynamics. The server shows potential allosteric sites and allows users to explore communication between the regulatory and functional sites. It is possible to explore, for instance, potential effector binding sites in a given structure as targets for allosteric drugs. As input, the server only requires a single structure. The server is freely available at http://allostery.bii.a-star.edu.sg/. - INTRODUCTION Protein function depends on the inherent dynamics of the protein structure. Not only is the balance between different conformational states of importance in this context, but also how easily the transitions between them occur. The external factors, such as ligand binding or local chemical modifications, can affect the conformational ensemble and shift the equilibrium toward (in)active conformations. The regulation is called allosteric when the effector site is not directly adjacent to the site of altered activity (1). The early phenomenological Monod-WymanChangeux (MWC) (2) and Koshland-Ne methy-Filmer (KNF) (3) models were devised to explain a classic example of allosteric regulation (4): the cooperative ligand binding of many oligomeric proteins, where binding of substrate to one subunit affects the ligand affinity in other identical subunits. The MWC model postulates that binding stabilizes one of several available conformations with emphasis on symmetry conservation, whereas the KNF model assumes an induced-fit scenario. Since the MWC and KNF models, numerous studies have been performed at different levels of coarsegraining (5). The models themselves have been expanded as well, and allostery is currently considered in proteins of different size, shape and degree of oligomerization, spanning from small single-domain structures to the large chaperones (6,7). Originally, there was an apparent dichotomy between MWC and KNF models and their counterparts in the energy landscape-based new view of allostery (811)conformational selection and induced fit. The main difference between the two models is whether binding precedes conformational change (11). Transition pathway analysis is primarily a matter of kinetics, whereas the shift in conformational equilibrium is one of thermodynamics: the conformational states involved determine which binding sites are allosterically connected, and their relative stability before and after binding determines the effect of regulation (12). Overall, the two models do, however, not describe mutually exclusive scenarios (6,11): in both cases, there is a shift in the population of different functional states on effector binding. The issue was resolved with the introduction of a more general physical framework (13). Despite the progress achieved in the understanding of allostery, most of studies have been performed on individual proteins or small collections of them (6,7,14). The previously developed approaches to the analysis of protein dynamics are mostly focused around the analysis of the energetics of the proteins structural ensemble, mobility of individual residues and conformational changes. For example, the COREX/BEST algorithm (15) enumerates the protein ensemble, defines the relative free energies of each state and characterizes the energetics of the ensemble. The AD-ENM server performs an analysis of macromolecular dynamics based on the calculation of the spectrum of normal modes for the elastic network model (16). The ProDy project allows to analyze dynamical properties of individual residues and to visualize protein dynamics (17). However, a general molecular description of allosteric regulation that allows prediction of allosteric sites based on protein dynamics, and that explains molecular mechanisms of communication between sites was still lacking (5). Resorting to the thermodynamic view of allostery (57), we developed the concepts of binding leverage and leverage coupling that allow quantifying (i) the coupling between ligand binding and the intrinsic dynamics of the protein and (ii) the communication between different binding sites. These concepts also allow finding latent effector binding sites, which along with known ones can be considered as potential targets for allosteric drugs (5). In the era of structural proteomics, with an exploding number of protein structures, it is of crucial importance to have instruments that allow massive and efficient analysis of multiple protein targets. For studying allostery, there are several important requirements for such an instrument. It should be based on a generic molecular model of allostery, which works regardless of the size, degree of oligomerization or function of the protein. It should work with a single structure, regardless of it corresponds to the active/activated or inactive/inactivated state of the protein. It should be able to explore communication between natural allosteric and catalytic sites, to detect latent sites in the structure, as well as to analyze sites chosen by the user. The SPACER server satisfies the aforementioned requirements, providing reasonably fast interactive tools for exploratory analysis of allosteric communication. Later in the text, we provide a brief description of the theoretical background for SPACERs methods followed by a practical guide to exploratory analysis of allosteric communication with SPACER. An online tutorial (http://allostery.bii.a-star.edu.sg/tutorial/) exemplifies the server workflow for the case of the Phosphofructokinase (PFK) homotetramer, showing the major options in the SPACER and explaining the most important features and results. THEORETICAL BACKGROUND The balance between different conformations of a protein and the role of ligand binding in switching between its functional states are the major determinants of allosteric regulation and communication. The steps in the analys (...truncated)